A computing device for generating and executing planned tasks based on semantic supervision and execution feedback.

Abstract

A computing device for managing planned execution tasks stores a planning agent model and a monitoring agent model. After obtaining an intent semantic related to a task of planned execution, the computing device uses the planning agent model to obtain a planned execution task plan, and then uses the monitoring agent model to determine whether the planned execution task plan conforms to the intent semantic. When the planned execution task plan is determined to conform to the intent semantic, the computing device obtains a task execution method to be checked based on the planned execution task plan, and uses the monitoring agent model to determine whether the task execution method to be checked passes the check. When the task execution method to be checked passes the check, the task execution method to be checked is treated as an executable method and the task is executed.

Claims

1. A computing device for generating and executing planned tasks based on semantic supervision and execution feedback, communicatively connected to a user device via a communication network, the computing device comprising: a communication unit communicatively connected to the user device via the communication network; a storage unit storing a planning agent model and a supervision agent model; and a processing unit electrically connected to the communication unit and the storage unit; wherein, The processing unit obtains an intent semantic from the user device related to a task to be executed. Based on the intent semantic, the processing unit uses the planning agent model to obtain a task execution plan. Based on the intent semantic and the task execution plan, the processing unit uses the supervision agent model to determine whether the task execution plan conforms to the intent semantic. When the processing unit determines that the task execution plan conforms to the intent semantic, it obtains a task execution method to be checked based on the task execution plan. Based on the intent semantic, the task execution plan, and the task execution method to be checked, the processing unit uses the supervision agent model to determine whether the task execution method to be checked passes the check. When the processing unit determines that the task execution method to be checked passes the check, it treats the task execution method to be checked as an executable method and executes the task according to the executable method.

2. The computing device for generating and executing planned execution tasks based on semantic supervision and execution feedback as described in claim 1, wherein, The intent semantics includes multiple task steps corresponding to the task. Based on the intent semantics and the planned execution task plan, the processing unit uses the supervisory agent model to determine whether the purpose of the planned execution task plan is consistent with the purpose of the intent semantics. When the processing unit determines that the purpose of the planned execution task plan is consistent with the purpose of the intent semantics, it determines whether the planned execution task plan is related to at least one key step among the task steps. When the processing unit determines that the planned execution task plan is related to the at least one key step, it determines that the planned execution task plan conforms to the intent semantics and obtains the task execution method to be checked.

3. The computing device for generating and executing planned execution tasks based on semantic supervision and execution feedback as described in claim 1, wherein the processing unit determines, based on the planned execution task plan and the execution method of the task to be checked, whether the purpose of the execution method of the task to be checked is consistent with the purpose of the planned execution task plan; when it is determined that the purpose of the execution method of the task to be checked is consistent with the purpose of the planned execution task plan, the processing unit determines, based on the intention semantics and the execution method of the task to be checked, whether the purpose of the execution method of the task to be checked is consistent with the intention semantics; when the processing unit determines that the purpose of the execution method of the task to be checked is consistent with the intention semantics, it determines that the execution method of the task to be checked has passed the check, and the execution method of the task to be checked is designated as the executable method, and the task is executed according to the executable method.

4. The computing device for generating and executing planned execution tasks based on semantic supervision and execution feedback as described in claim 1, wherein the processing unit generates a task plan to be confirmed corresponding to the intent semantics and a task planning dialogue script corresponding to the task plan to be confirmed using the planning agent model according to the intent semantics; the processing unit obtains and transmits at least one task-oriented dialogue to the user device according to the task planning dialogue script to obtain at least one user response information; and the processing unit obtains the planned execution task plan according to the at least one user response information and the task plan to be confirmed.

5. The computing device for generating and executing planned execution tasks based on semantic supervision and execution feedback as described in claim 1, wherein the processing unit obtains an execution log containing the execution result based on an execution result generated by executing the task corresponding to the executable method, the processing unit determines whether the execution result is a successful result based on the execution log, and when the processing unit determines that the execution result is a successful result, stores the executable method in the computing device.

Images